speed up DA clusterizer allowing more vectorization

[VinInn]

35% speed up in the DA clusterizer itself at low pileup.
small regression observed in primary vertex reconstruction and other products whose producer algorithms depend on primary vertices (detached tracks, conversion)

https://twiki.cern.ch/twiki/bin/viewauth/CMS/HighGranularityCMSSWOptimization

[cmsbuild]

A new Pull Request was created by @VinInn (Vincenzo Innocente) for CMSSW_7_2_X.

speed up DA clusterizer allowing more vectorization

It involves the following packages:

Alignment/OfflineValidation
RecoVertex/PrimaryVertexProducer

@nclopezo, @cmsbuild, @diguida, @rcastello, @StoyanStoynev, @slava77, @Degano can you please review it and eventually sign? Thanks.
@ghellwig, @frmeier, @GiacomoSguazzoni, @rovere, @cerati, @mmusich this is something you requested to watch as well.
You can sign-off by replying to this message having '+1' in the first line of your reply.
You can reject by replying to this message having '-1' in the first line of your reply.

[cmsbuild]

Pull request #4435 was updated. @nclopezo, @cmsbuild, @Degano, @StoyanStoynev, @slava77 can you please check and sign again.

[cmsbuild]

+1
Tested at: 37e4476
https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-4435/208/summary.html

[cmsbuild]

Comparison is ready
https://cmssdt.cern.ch/SDT/jenkins-artifacts/pull-request-integration/PR-4435/208/summary.html

[VinInn]

how could be that
https://cmssdt.cern.ch/SDT/jenkins-artifacts/baseLineComparisons/CMSSW_7_2_X_2014-06-30-0200+4438/2987/25.0_TTbar+TTbar+DIGI+RECO+HARVEST+ALCATT/HLT_Tracking_iter4Merged_GeneralProperties.html
looks identical to
https://cmssdt.cern.ch/SDT/jenkins-artifacts/baseLineComparisons/CMSSW_7_2_X_2014-06-30-0200+4435/2983/25.0_TTbar+TTbar+DIGI+RECO+HARVEST+ALCATT/HLT_Tracking_iter4Merged_GeneralProperties.html

???

[StoyanStoynev]

Working on it (for the above and for the record - see #4438).

[StoyanStoynev]

I'll have it finalized by tomorrow. At least I can say I have no issues with the code changes. There are a lot of (seemingly expected) differences seen in QCD and ttbar (with PU) samples that I need to sort out - all look a fluctuation ("random") type, no trend, acceptable. This is to finalize and also check timing.

[StoyanStoynev]

Timing. Tested on wf 202 (ttbar with PU; 200 events) and wf 38 (QCD, 1000 events). wf 25202 is consistent but has somewhat lesser precision (less events and less time per event for the interesting quantities). Overall - vertex times are affected positively (upto ~ 30+ %), some secondary vertex related quantities are affected negatively.

wf 202 (note: should apply +10% for proper normalization):
pixelVertices 55.8764 ms/ev -> 39.6211 ms/ev (after normalization would be ~ 44 ms/ev)
firstStepPrimaryVertices 70.3883 ms/ev -> 54.6836 ms/ev
offlinePrimaryVertices 144.392 ms/ev -> 111.15 ms/ev

negatively affected:
combinedSecondaryVertexBJetTags 2.12806 ms/ev -> 2.67696 ms/ev

not affected within few % (for the record):
particleFlowDisplacedVertex 57.2981 ms/ev -> 52.3893 ms/ev
secondaryVertexTagInfos 4.79849 ms/ev -> 4.28263 ms/ev
ghostTrackVertexTagInfos 9.97337 ms/ev -> 8.97076 ms/ev
hpsPFTauPrimaryVertexProducer 59.2892 ms/ev -> 55.452 ms/ev

In wf 38 (no PU) gains are much less pronounced (note: should apply +15% for proper normalization):
pixelVertices 29.1288 ms/ev -> 25.1571 ms/ev
firstStepPrimaryVertices 49.6234 ms/ev -> 42.6784 ms/ev
offlinePrimaryVertices 77.8378 ms/ev -> 65.9154 ms/ev

combinedSecondaryVertexBJetTags 4.25616 ms/ev -> 7.02886 ms/ev

particleFlowDisplacedVertex 16.0216 ms/ev -> 13.4798 ms/ev
secondaryVertexTagInfos 15.4817 ms/ev -> 13.0882 ms/ev
ghostTrackVertexTagInfos 36.6605 ms/ev -> 30.2541 ms/ev

Event Size. It is affected including other objects further down the line but in a mild way
(wf 202 tested). Examples are given below to illustrate changes.

"Branch Name | Average Uncompressed Size (Bytes/Event)":
recoVertexs_offlinePrimaryVertices__RECO. 14984.7 -> 14978.7
recoVertexs_pixelVertices__RECO. 8066.63 -> 8042.84
recoSecondaryVertexTagInfos_secondaryVertexTagInfos__RECO. 4137.22 -> 4139.88
recoTracks_generalTracks__RECO. 282111 -> 282031
recoJetedmRefToBaseProdTofloatsAssociationVector_combinedSecondaryVertexBJetTags__RECO. 380.555 -> 380.775
recoPFDisplacedVertexs_particleFlowDisplacedVertex__RECO. 616.42 -> 616.42 (not affected)
recoPFTaus_hpsPFTauProducer__RECO. 9980.17 -> 9976.24

Comparison plots with summary will follow shortly.

[StoyanStoynev]

Comparisons. The number of PV (and constituents) is affected and this propagates to affect distributions in other objects (low and high level). Most of the changes are practically invisible. Taus and parts of isolation are probably the most significant (still not large but visible) to mention.

PV:

Tau:

Muon Iso /note some shapes are affected; the only significant shape change i spotted/:

Jet/MET:

Low level:

All differences are acceptable and most are either negligible or not affecting the actual shapes of the distributions (based on this statistics but I see no hints for the opposite).

[StoyanStoynev]

(red above is this PR)

[StoyanStoynev]

+1
Based on the review above. In addition - most of the code changes are editorial (indentations), the non-trivial ones are optimizations (plus some additions). A note: there are new warnings appearing (wf 202):

%MSG-w TwoTrackMinimumDistance: PFTauPrimaryVertexProducer:hpsPFTauPrimaryVertexProducer 02-Jul-2014 21:03:13 CEST Run: 1 Event: 5088
comparing track with itself!
(---and---)
Computation HelixHelix::CAIR failed.

It happened in one event only (though 77 times) out of 200 events.

Tested 37e4476 on top of CMSSW_7_2_X_2014-06-27-0200 (as #4437, #4438, #4447). It is still mergeable in CMSSW_7_2_X_2014-07-04-0200 .

[cmsbuild]

This pull request is fully signed and it will be integrated in one of the next CMSSW_7_2_X IBs unless changes (tests are also fine).